HxA partners with AFA and FIRE to call for institutional neutrality.

College trustees: put truth-seeking above politicized institutional statements.

Learn more.
Heterodox Academy
Back to Blog
Mm free speech
November 16, 2023+Michael Muthukrishna
+Viewpoint Diversity+Campus Climate+Public Policy

Innovation and Social Progress Require Unfettered Free Speech

The following is an excerpt from A Theory of Everyone: The New Science of Who We Are, How We Got Here, and Where We're Going, by Michael Muthukrishna. Published by The MIT Press. Copyright © 2023 Salgado Muthukrishna Consulting Ltd. All rights reserved.

Language, speech, talking to one another. These are social synapses that are firing in our collective brain. Restricting that firing reduces trust in one another and cripples our capacity for innovation. It is collective brain damage.

In his vigorous defense of liberty, philosopher John Stuart Mill reminded us ‘that it is important to give the freest scope possible to uncustomary things, in order that it may in time appear which of these are fit to be converted into customs’. Mill was describing what we would now call cultural evolution – the evaluation of different ideas and the spread of those that work best. But cultural evolution is most efficient when ideas are allowed to flow freely. Limitations on free speech are like blockages in the pipes of progress, preventing us from seeing the world as it is. And if we can’t see the world as it is, we can’t figure out how to fix it.

We used to believe the Earth was the center of the universe. You got into trouble for suggesting otherwise. We used to believe that it was natural that some people should be able to own others. Wars were fought to end slavery. For the uncustomary to become customary, we must be exposed to alternatives so that we can evaluate them.

To solve problems, we must first understand them. Take the gender wage gap for example – the gap between the average earnings of all women versus all men, often quoted as 70 cents to the 1 dollar. If we stubbornly insist that the main explanation for this gap is bias, discrimination, and/or sexism, we are guilty of either not being specific enough about what constitutes and causes bias, sexism, and discrimination to be able to act, or, worse still, we may jump to solutions that may not work or, like proximate patches, create new problems. The large industry of HR implicit bias training or attempts to debias hiring committees are examples of failed policies that lack evidence. A 2019 review of 492 implicit bias studies consisting of 87,418 participants found no evidence that implicit bias training programs brought about behavioral change. Other reviews suggest diversity training can even backfire, increasing bias.

An alternative approach is a free exchange of hypotheses with critical tests and remediations based on evidence. Seventy cents to the dollar is a statistic that looks at the median wage of all men and all women across all occupations – the unadjusted wage gap. It compares high-paid majority male CEOs with low-paid majority female social workers. What leads men and women to these different careers with different tasks, wages, and lifestyles may be a result of limited choice, boys and girls receiving different encouragement to pursue different careers, differences in long-standing evolved preferences for types of tasks, inequalities in male and female participation in child-rearing, lack of societal support for balancing child-rearing and careers, or many other possibilities. When related factors such as occupation, experience, or hours worked are controlled – the adjusted wage gap – the size of the gap shrinks or disappears. But these controls, regardless of what they are, are precisely the policy levers and potential solutions we need to address the wage gap.

Easy answers or certainty over what the answer must be – or worse still, sanctioning unfavorable hypotheses – hinders our ability to discover the truth and act on it.

"We can only arrive at the truth in a diverse environment of different backgrounds, considering all hypotheses and ideas – both those we like and those we don’t."

The world is a complicated place and doesn’t always conform to what we think or hope the answers are. We can only arrive at the truth in a diverse environment of different backgrounds, considering all hypotheses and ideas – both those we like and those we don’t. Cultural evolution needs the fuel of diversity and free speech to create the variation for transmission and selection. This kind of exploration needs a diversity of people with different experiences to come together in a safe space that enables unfettered free speech.

This is how science works best.

We are all biased. We all think we’re on the factually and morally correct side of any debate. None of us is immune and none of us can even see the full extent of our bias – what’s called the bias blindspot. I’d like to think that scientists are more likely to change their minds in the face of new evidence, but I know that this is rare even among scientists trying their hardest to be unbiased.

Science doesn’t work because we’re enlightened humans who see past our incentives and our life experience. It doesn’t work because we readily change our minds in the face of new evidence. No, science works because we commit to a method of discovery, there is agreement on what counts as evidence, and, most importantly, we are incentivized to show others that they’re wrong. It’s a collective act that slowly converges on the truth. But our findings can only be trusted if we are free to find the opposite to whatever current political sentiments suggest is the right answer.

In 2020 researchers Bedoor AlShebli, Kinga Makovi, and Talal Rahwan published a paper in Nature Communications, a journal within the prestigious Nature family of academic journals. The paper was titled ‘The Association between Early Career Informal Mentorship in Academic Collaborations and Junior Author Performance’. Using the Microsoft Academic Graph database of scientific papers, citation networks, and information about authors, they found that the work of female scientists with female mentors is associated with lower scientific impact than female scientists who have male mentors. They argued that this finding raises ‘the possibility that opposite-gender mentorship may actually increase the impact of women who pursue a scientific career. These findings add a new perspective to the policy debate on how to best elevate the status of women in science.’

Understandably, the paper was met with a swift negative response on social media. There is a real potential that these findings could lead to female researchers receiving fewer applications from talented female students, a feedback loop that could further perpetuate whatever might be driving the finding (assuming the finding replicates). The negative response eventually led to a retraction. With no further context, this might have been the story of a bad paper with careless consideration for the implications of the findings. But it is instead the story of how bias can creep into the scientific record, distorting truth and affecting our ability to understand the world.

In 2018, just two years earlier, Nature Communications had published another paper, this time by Bedoor AlShebli, Talal Rahwan, and Wei Lee Woon. This was titled ‘The Preeminence of Ethnic Diversity in Scientific Collaboration’. Using the same Microsoft Academic Graph database, the authors had used a similar method, finding that ‘ethnic diversity had the strongest correlation with scientific impact. To further isolate the effects of ethnic diversity, we used randomized baseline models and again found a clear link between diversity and impact.’ They concluded that ‘recruiters should always strive to encourage and promote ethnic diversity’.

This paper, published by two of the same authors in the same journal with the same dataset and similar methods, was met with praise and remains, unretracted, in the scientific record.

As with any paper, both studies have their strengths and flaws, and supporters and critics can point to these as evidence for why one, both, or neither deserved to be retracted or never published.

The main critique of work like this often boils down to perceived harmful effects. A finding from the same dataset that female mentors are bad for female scientists is harmful, but a finding that ethnic diversity is good for impact is beneficial. This harm and benefit may very well be true, but there is a long-term cost to this short-term thinking. Selective condemnation based not on accuracy but harm supports what Plato called a noble lie – something false that is nonetheless maintained and promoted because it has perceived positive effects on society. Noble lies, especially when naked in their bias, squander scientific credibility. Whatever the truth, if people don’t believe that for any finding an opposite finding could have also been published, then science becomes nothing more than untrustworthy propaganda, mirroring and supporting what we already believe.

"Many point out that there are problems associated with a policy of free speech, but these problems are not resolved by restricting speech."

The solution to misinformation is more information. The answer to the infamous fire in a theatre analogy is that when someone falsely shouts ‘Fire!’, we need other voices to shout ‘No there isn’t!’ And we need to exploit indirect reciprocity, tracking reputations like the boy who cried wolf, turning false fire alarmists into untrustworthy sources who lack credibility in other domains of life.

Many point out that there are problems associated with a policy of free speech, but these problems are not resolved by restricting speech. For example, some fear that power differences mean that speech is never truly free because some voices will be louder than others. But this problem is only made worse by restricting free speech. Powerful voices that can shout loudly in an environment of free speech can be balanced by whispers that grow into roars. But in an environment where speech is restricted, those same powerful voices can ensure alternative softer voices never speak at all.

This is not a left or right issue. There are real legislative proposals restricting speech in both directions. In US states governed by Democrats laws have been proposed to regulate social media companies when it comes to hate speech and extremism. In US states governed by Republicans laws have been proposed to prevent social media companies from removing hate speech. Scientists are disincentivized from saying things that would upset donors or their funders or cause them to lose their livelihoods. Some have attempted to pass laws that explicitly restrict what scientists can say. This is a bipartisan issue that affects all of us. None of us likes hearing things we disagree with, things that undermine our livelihoods or challenges features of the world that benefit us. If you have enough information to audit people’s incentives, you’ll often find that they advocate for things that directly or indirectly improve their material well-being, reputation, or status.

Unfettered free speech is historically unusual. That’s why it was explicitly enshrined in the First Amendment of the US Constitution. It was an unusual and important cultural innovation in a world where laws prosecuted blasphemy, offensive speech, or insulting the monarch. Many of these laws are still on the books in the old worlds of Asia and even Europe. In 2009 an Austrian woman was fined for insulting the Prophet Mohammad, a ruling upheld in 2011 by the European Court of Human Rights. In 2022, after the death of Queen Elizabeth II, British anti-monarchist protesters with Not My King placards were threatened with arrest. British police regularly arrest people over online posts and comments deemed racist, offensive, or hate speech – not protected speech in the UK. Where the freedom to speak is uncertain, it has a chilling effect on people’s willingness to express their honest opinions and the free exchange of the ideas that lead to innovation and creativity.

Of course, the US First Amendment only restricts government, but the broader principle of free and open exchange – free speech – protecting even and especially speech we would rather see banned, is essential to science and progress. The First Amendment has bled over into a culture of free expression that may be the reason that some of the biggest social movements of the last 100 years from the sexual revolution to Silent Spring to Black Lives Matter and #MeToo began in America. Free speech culture now requires protection.

The importance of a safe space to speak one’s mind and where we can reveal how we think the world really works is emphasized by research on psychological safety.

In a famous analysis conducted by Google, Project Aristotle, psychological safety – the freedom to express what you really think and resilience to hearing ideas you don’t like – emerged as a key prerequisite for successful teams, especially diverse teams. Diverse minds can only combine into a brilliant collective brain when they trust one another. When people trust each other, they feel more comfortable saying what they really think. And that’s critical to being shown why they’re wrong, showing others why they’re wrong, or discovering a new truth at the intersection of ideas and beliefs previously isolated in the heads of different people.

We can create a culture of psychological safety by encouraging in interpersonal interactions what’s called the robustness principle (Postel’s Law) in software engineering: be conservative in what you do, be liberal in what you accept from others. In programming this refers to, for example, outputting well-formatted files but robustly accepting poorly formatted files from other programs. Don’t create files with missing sections or stray semicolons, but be able to read them if they have these mistakes. In communication it means being generous to others’ intentions – ‘steel-manning’ rather than ‘straw-manning’ their position – but personally making an effort to communicate clearly. A steel man argument attempts to make the best version of someone else’s argument, perhaps even better than they made themselves. Steel-manning is the strategy for people who see arguments as a means of together arriving at the truth rather than sparring to win for the sake of winning. Straw-manning is the opposite, taking the weakest version of an argument and arguing against it with the goal of winning rather than learning.

It’s hard to know in advance what will and what won’t cause harm in the long term, and so the bar for what we should ban must be incredibly high and err on the side of not banning or suppressing speech. This is particularly important when it comes to freedom of speech in science, because reality is complicated and science is never settled. If we can’t trust that the scientists are speaking freely then how can we trust science? Take, for example, group differences and the role of genes. Based on my reading of the evidence, I come down on the side of culture being the primary cause of these differences, but you should only believe me if those who believe otherwise are equally able to express the most defensible version of their position. If the only voice that could be heard in polite circles was mine then how could you know that there was an alternative argument and hear the evidence for it?

"No one has a right or even a reasonable expectation to not be offended. We make progress not by bludgeoning our opponents into believing they are bigots and bad people, but by finding out what people think and why they think it."

No one has a right or even a reasonable expectation to not be offended. We make progress not by bludgeoning our opponents into believing they are bigots and bad people, but by finding out what people think and why they think it. What evidence they have and what critical evidence they would need to change their mind. The reasons why we disagree with someone have to be based on logic and evidence, not group membership or prejudices. To err is human, but being wrong should not end people’s livelihoods or their ability to say something else. Though it may change the credibility of what they say, it should not change their ability to say it. It would be like preventing an entrepreneur from starting another company because their previous company had failed. So many innovations would be lost in such a closed culture. Free speech is critical to the process of discovery and the triggering of a creative explosion, especially for social innovation. The vibrancy of the United States – its robust debates, fights and protests, and even racist and anti-racist polarizing discussions – is a product of the freedom to speak freely. But the goal should not be to win the argument, it should be to arrive at the truth as a society. The norms of how we argue are critical to this process.

The world is a complicated place. Yesterday’s obvious truths are today’s falsehoods and vice versa. As scientists we try our best to steel-man an opposition’s arguments. The goal is not to win the argument, but to arrive at the truth. And that means seeing our opponent not as an opponent but as a fellow truth seeker who deserves whatever assistance we may render in developing the best version of their argument no matter how wrong we think it is.

That’s why censorship by government institutional misinformation tribunals or making Mark Zuckerberg, Elon Musk, or whomever they designate arbiter over what is true and false is not a solution to the real problem of misinformation. Instead, like many proximate, band-aid solutions, it creates new problems. Misinformation is a real challenge, but banning or suppressing speech is not the answer.

The person in charge of deciding the truth may not always be the person you want. Remember the opposition may one day be in power and a person with different politics may one day run the company. Laws must be made in a Rawlsian manner, prepared for a time when you and yours are no longer in charge. We must be ruled by principles, not people. For tackling misinformation, we want emergent, systems-level solutions such as offering more, not less, context and information.

There is value in minimizing the ability to create bots and fake accounts that poison our social network streams by flooding them with content that is not representative of what actual people think in an attempt to exploit our social learning biases. But real people should not be restricted in what they have to say.

A lack of free speech cripples the collective brain. It deprives it of information, the sparks of ideas moving from person to person that could light the next fire that radically changes our lives.


Get HxA In Your Inbox

Related Articles
Abhishek Saha Protecting free speech in universities
Protecting Free Speech in Universities: Insights from the UK
February 14, 2024+Abhishek Saha
+Viewpoint Diversity+Open Inquiry+Campus Policy+Campus Climate
Make a donation
Make a Donation

Your generosity supports our non-partisan efforts to advance the principles of open inquiry, viewpoint diversity, and constructive disagreement to improve higher education and academic research.

This site use cookies.

To better improve your site experience, we collect some data. To see what types of information we collect, read our Cookie Policy.